提交 87de31bf 编写于 作者: Y Yibing Liu

update doc information

上级 7c423e4b
......@@ -48,9 +48,9 @@ class RankLossOpMaker : public framework::OpProtoAndCheckerMaker {
RankLossOpMaker(framework::OpProto *proto,
framework::OpAttrChecker *op_checker)
: OpProtoAndCheckerMaker(proto, op_checker) {
AddInput("P", "The first input of RankLoss operator.");
AddInput("Oi", "The second input of RankLoss operator.");
AddInput("Oj", "The third input of RankLoss operator.");
AddInput("P", "The desired target values for posteriors.");
AddInput("Oi", "The model output for item i.");
AddInput("Oj", "The model output for item j.");
AddOutput("Out", "The output tensor of RankLoss operator.");
AddComment(R"DOC(RankLoss operator
......@@ -63,6 +63,8 @@ three inputs: P, Oi, and Oj, and the rank cost can be expressed as
\tilde{P_{i,j}} = \left \{0, 0.5, 1 \right \} \ or \ \left \{0, 1 \right \}
\f]
A detailed explanation about these notations can be found in
[1]. Chris Burges, Tal Shaked, Erin Renshaw, et al. Learning to
Rank useing Gradient Descent.
)DOC");
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册